首页|"弱强超"AI的仿人理路及风控演化研究

"弱强超"AI的仿人理路及风控演化研究

扫码查看
AI核心的研发思路和演进原则是仿人,且AI的智能层次与智能水平及价值风控因"弱强超"三级智能AI仿人的功能或程度不同而有所差异.弱AI主要模仿人的局部智能,仿人的风控设置简单、方式单一;强AI或通用AI模仿人的内容复杂,仿人的风控理路多元、集成性强,其核心是要求AI遵循人的一般正常思维和行为准则;超AI阶段,既要确保智能机器在正常情况下的安全、可靠、可控、可用,又要确保在异常情况下有理性的干预与价值调控.对仿人原则的落实,既要求在AI研发应用的全过程中严格遵循人的一般正常思维和行为价值准则,又要求AI植入道德智能系统和价值智能系统,即为了确保三级智能AI始终围绕造福人类的价值指向健康持续发展,势必要把AI的智能水平和仿人原则实现的风控条件及演进规律放在AI技术层面、伦理层面等多方面、多维度地进行深化研究.
The Humanoid Path of"Weak,Strong,and Super"AI and the Evolution of Risk Control
AI core R&D ideas and evolutionary principles is to imitate people,and AI's intelligence leveland value of wind control is different due to the"weak strong super"three-level intelligence AI imitation function ordegree.Weak AI mainly imitates the local intelligence of human beings,and the wind control of imitating human beings is simple and single;strong AI or general AI imitates human beings with complex content,and the wind control of imitating human beings is diversified and integrated,and the core of it is to require AI to follow the normal thinking and behavioral norms of human beings;at the stage of super AI,it is not only necessary to ensure that the intelligent machines are safe,reliable,controllable,and usable under normal circumstances,but also to ensure that there is a rational intervention and value regulation under abnormal circumstances.The implementation of the humanoid principle not only requires that the general normal human thinking and behavioral value guidelines be strictly followed in the whole process of AI R&D and application,but also requires that the CPU of the AI implant the ethical intelligence system and the value intelligence system,in order to ensure that the three-stage intelligent AI is always centered on the healthy and sustainable development of value pointing to the benefit of mankind,the level of the intelligence of the AI and the wind-control conditions for the realization of the humanoid principle and the law of evolution are inevitably placed on the AI technology level,ethical level and other multi-faceted,multi-dimensional deepening research.

潘军

展开 >

贵州商学院

清华大学

中国科学院大学

人工智能 智能水平 仿人原则 理性价值 价值风控

国家社会科学基金一般项目贵州省科协重大项目

21BKS165GZKX2023ST002

2024

阅江学刊
南京信息工程大学

阅江学刊

CHSSCD
影响因子:0.313
ISSN:1674-7089
年,卷(期):2024.16(4)
  • 33